Suraj Karakulath

I am an AI Engineer/Data scientist with a background in marketing. I started off in copywriting, then grew into “content” and strategy development, while also handling data analytics and measurement. Over 8 years, I got to work with some leading tech enterprises and startups specialising in computer vision, artificial intelligence, geospatial data analytics and conversational AI/chatbot development, which inspired me to formally move into Data Science.

I grew up in India, lived and worked in Singapore for 16 years and now I’m based in Germany, working for DEPT, a global technology and marketing agency.

My specialties include:

• AI Engineering and Agentic workflow orchestration
• Traditional ML and end-to-end system development
• Deep learning and NLP use cases
• Time series analysis and forecasting
• Data Science practice - covering mining, exploration, statistical modeling and visualisation

Tools/skills that I work with:

• Agentic AI Engineering - Claude, Cursor, Codex
• LLM APIs - Vertex AI, Google AI Studio, Claude API and OpenAI API
• IDE: Antigravity, Visual Studio, RStudio, Cursor
• Cloud deployment environments: GCP primarily, Azure occasionally
• Database: SQL, BigQuery, Snowflake
• Languages: Python for general, Typescript and Reach for frontend, R for data science sometimes
• General data science and ML: pandas, numpy, scikit-learn, statsmodels, NLTK, Spacy, XGBoost,
• Deep learning: PyTorch, Tensorflow
• NLP: HuggingFace transformers, OpenAI GPT models, Claude API, LangChain, LlamaIndex
• Version control and collab: Git, GitHub
• Quick dashboarding and prototyping: Streamlit, FastHTML
• BI tools: Tableau, Google Data Studio (Looker Studio), PowerBI
• Web analytics: Google Analytics, Search Console, Adobe Analytics, HubSpot

Some projects that I did for work include

• a product clustering and constrained optimization use case during my time at rebuy, a recommerce startup in Berlin
• an ETL pipeline for clustering protein binding sites for a university research team
• a time series forecasting study in non-transparent markets using ARIMA, XGBoost and Prophet for a market-leading energy supplier in Germany
• an unsupervised clustering of website audiences into different segments, and
• an attempt at quantifying global temperature rise from climate change data </a>.

Some sideprojects that I had fun with include

• an interactive visualisation of audio features of The Beatles songs using Spotify API,
• another interactive visualisation of characters’ sentiment from the scripts of the Star Trek (original) series and
• an analysis to detect political bias in ChatGPT responses using NLP.

My other interests include technology, cinema and film history, music, politics, psychology and science.

When I’m not working, I’m usually

• travelling (mostly Europe now since I am here)
• reading (non-fiction - psychology, science, humanism; fiction - classics)
• watching films (classics - a bit obsessed with Danny Boyle and Aaron Sorkin these days) or checking out what’s on at local arthouse theatres
• learning German
• or catching up with friends